[
https://issues.apache.org/jira/browse/SPARK-56948?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Kent Yao resolved SPARK-56948.
------------------------------
Fix Version/s: 5.0.0
Resolution: Fixed
Issue resolved by pull request 55988
[https://github.com/apache/spark/pull/55988]
> TPCDSQueryBenchmark: make hardcoded SparkConf overridable via setIfMissing
> --------------------------------------------------------------------------
>
> Key: SPARK-56948
> URL: https://issues.apache.org/jira/browse/SPARK-56948
> Project: Spark
> Issue Type: Improvement
> Components: Tests
> Affects Versions: 5.0.0
> Reporter: Kent Yao
> Assignee: Kent Yao
> Priority: Major
> Labels: pull-request-available
> Fix For: 5.0.0
>
>
> Currently TPCDSQueryBenchmark hardcodes spark.driver.memory=3g,
> spark.executor.memory=3g, and spark.sql.autoBroadcastJoinThreshold=20MB via
> .set(...), which takes precedence over -Dspark.* system properties.
> This prevents tuning heap/broadcast when running the benchmark at larger
> scale factors (SF10 / SF100 / SF1000), where the 3g default OOMs.
> spark.sql.shuffle.partitions already supports override via System.getProperty
> in the same file -- extend the same pattern (via setIfMissing for
> consistency) to the remaining three keys. Defaults unchanged, fully backward
> compatible.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
